Paweł Pacana

@msencenb I've managed to recreate the issue you're experiencing: https://gist.github.com/pawelpacana/f39026514185e8f72e4daaca2d99b85a I'll look into it more in the upcoming days. It seems that `PreserveTypes` transformation used by default in `JSONClient` is...

@msencenb There's a workaround which might work without introducing too much changes: ``` preserve_types_as_seen_in_json_client.register(OpenStruct, serializer: ->(v) { YAML.dump(v) }, deserializer: ->(v) { YAML.unsafe_load(v) }), ``` That is using YAML only...

A bit stale already and no momentum to push this forward. Closing, but not discouraging.

Just wanted to say I'm revisiting this PR now. I've re-read linked articles and the comments you've all left in this thread. Summary of all mentioned problems with SAVEPOINT: -...

> Are you able to explain why you are using the subtransaction in the first place? I don't think there's a good reason to use subtransaction in RES as it...

I'll rebase this PR next and review existing tests around transactions and expectations [described earlier](https://github.com/RailsEventStore/rails_event_store/pull/1108#issuecomment-916375489). In the meantime I've [compared](https://github.com/RailsEventStore/rails_event_store/commit/c39132a0483e0daccce60101db3ffc7eda77554d) use-vs-no-use of SAVEPOINT on Postgres 15.5 and the difference is...

Already testing.

Any idea how to create/where to host reproducible benchmark environment? In other case effort is not justified.

So you want to have numbers that cannot be compared. Why bother than?

Ok I have misunderstood it based on too general title of this issue.